icmp6.c revision 1.189
1/* $NetBSD: icmp6.c,v 1.189 2016/06/21 10:25:27 ozaki-r Exp $ */ 2/* $KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 jinmei Exp $ */ 3 4/* 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33/* 34 * Copyright (c) 1982, 1986, 1988, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)ip_icmp.c 8.2 (Berkeley) 1/4/94 62 */ 63 64#include <sys/cdefs.h> 65__KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.189 2016/06/21 10:25:27 ozaki-r Exp $"); 66 67#ifdef _KERNEL_OPT 68#include "opt_inet.h" 69#include "opt_ipsec.h" 70#endif 71 72#include <sys/param.h> 73#include <sys/systm.h> 74#include <sys/kmem.h> 75#include <sys/mbuf.h> 76#include <sys/protosw.h> 77#include <sys/socket.h> 78#include <sys/socketvar.h> 79#include <sys/time.h> 80#include <sys/kernel.h> 81#include <sys/syslog.h> 82#include <sys/domain.h> 83#include <sys/sysctl.h> 84 85#include <net/if.h> 86#include <net/route.h> 87#include <net/if_dl.h> 88#include <net/if_types.h> 89 90#include <netinet/in.h> 91#include <netinet/in_var.h> 92#include <netinet/ip6.h> 93#include <netinet6/ip6_var.h> 94#include <netinet6/ip6_private.h> 95#include <netinet/icmp6.h> 96#include <netinet6/icmp6_private.h> 97#include <netinet6/mld6_var.h> 98#include <netinet6/in6_pcb.h> 99#include <netinet6/nd6.h> 100#include <netinet6/in6_ifattach.h> 101#include <netinet6/ip6protosw.h> 102#include <netinet6/scope6_var.h> 103 104#ifdef IPSEC 105#include <netipsec/ipsec.h> 106#include <netipsec/key.h> 107#endif 108 109 110#include "faith.h" 111#if defined(NFAITH) && 0 < NFAITH 112#include <net/if_faith.h> 113#endif 114 115#include <net/net_osdep.h> 116 117extern struct domain inet6domain; 118 119percpu_t *icmp6stat_percpu; 120 121extern struct inpcbtable raw6cbtable; 122extern int icmp6errppslim; 123static int icmp6errpps_count = 0; 124static struct timeval icmp6errppslim_last; 125extern int icmp6_nodeinfo; 126 127/* 128 * List of callbacks to notify when Path MTU changes are made. 129 */ 130struct icmp6_mtudisc_callback { 131 LIST_ENTRY(icmp6_mtudisc_callback) mc_list; 132 void (*mc_func)(struct in6_addr *); 133}; 134 135LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks = 136 LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks); 137 138static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL; 139extern int pmtu_expire; 140 141/* XXX do these values make any sense? */ 142static int icmp6_mtudisc_hiwat = 1280; 143static int icmp6_mtudisc_lowat = 256; 144 145/* 146 * keep track of # of redirect routes. 147 */ 148static struct rttimer_queue *icmp6_redirect_timeout_q = NULL; 149 150/* XXX experimental, turned off */ 151static int icmp6_redirect_hiwat = -1; 152static int icmp6_redirect_lowat = -1; 153 154static void icmp6_errcount(u_int, int, int); 155static int icmp6_rip6_input(struct mbuf **, int); 156static int icmp6_ratelimit(const struct in6_addr *, const int, const int); 157static const char *icmp6_redirect_diag(struct in6_addr *, 158 struct in6_addr *, struct in6_addr *); 159static struct mbuf *ni6_input(struct mbuf *, int); 160static struct mbuf *ni6_nametodns(const char *, int, int); 161static int ni6_dnsmatch(const char *, int, const char *, int); 162static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *, 163 struct ifnet **, char *); 164static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *, 165 struct ifnet *, int); 166static int icmp6_notify_error(struct mbuf *, int, int, int); 167static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *); 168static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *); 169static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *); 170static void sysctl_net_inet6_icmp6_setup(struct sysctllog **); 171 172 173void 174icmp6_init(void) 175{ 176 177 sysctl_net_inet6_icmp6_setup(NULL); 178 mld_init(); 179 icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire); 180 icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout); 181 182 icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS); 183} 184 185static void 186icmp6_errcount(u_int base, int type, int code) 187{ 188 switch (type) { 189 case ICMP6_DST_UNREACH: 190 switch (code) { 191 case ICMP6_DST_UNREACH_NOROUTE: 192 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE); 193 return; 194 case ICMP6_DST_UNREACH_ADMIN: 195 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN); 196 return; 197 case ICMP6_DST_UNREACH_BEYONDSCOPE: 198 ICMP6_STATINC(base + 199 ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE); 200 return; 201 case ICMP6_DST_UNREACH_ADDR: 202 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR); 203 return; 204 case ICMP6_DST_UNREACH_NOPORT: 205 ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT); 206 return; 207 } 208 break; 209 case ICMP6_PACKET_TOO_BIG: 210 ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG); 211 return; 212 case ICMP6_TIME_EXCEEDED: 213 switch (code) { 214 case ICMP6_TIME_EXCEED_TRANSIT: 215 ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT); 216 return; 217 case ICMP6_TIME_EXCEED_REASSEMBLY: 218 ICMP6_STATINC(base + 219 ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY); 220 return; 221 } 222 break; 223 case ICMP6_PARAM_PROB: 224 switch (code) { 225 case ICMP6_PARAMPROB_HEADER: 226 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER); 227 return; 228 case ICMP6_PARAMPROB_NEXTHEADER: 229 ICMP6_STATINC(base + 230 ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER); 231 return; 232 case ICMP6_PARAMPROB_OPTION: 233 ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION); 234 return; 235 } 236 break; 237 case ND_REDIRECT: 238 ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT); 239 return; 240 } 241 ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN); 242} 243 244/* 245 * Register a Path MTU Discovery callback. 246 */ 247void 248icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *)) 249{ 250 struct icmp6_mtudisc_callback *mc; 251 252 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 253 mc = LIST_NEXT(mc, mc_list)) { 254 if (mc->mc_func == func) 255 return; 256 } 257 258 mc = kmem_alloc(sizeof(*mc), KM_SLEEP); 259 mc->mc_func = func; 260 LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list); 261} 262 263/* 264 * A wrapper function for icmp6_error() necessary when the erroneous packet 265 * may not contain enough scope zone information. 266 */ 267void 268icmp6_error2(struct mbuf *m, int type, int code, int param, 269 struct ifnet *ifp) 270{ 271 struct ip6_hdr *ip6; 272 273 if (ifp == NULL) 274 return; 275 276 if (m->m_len < sizeof(struct ip6_hdr)) { 277 m = m_pullup(m, sizeof(struct ip6_hdr)); 278 if (m == NULL) 279 return; 280 } 281 282 ip6 = mtod(m, struct ip6_hdr *); 283 284 if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0) 285 return; 286 if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0) 287 return; 288 289 icmp6_error(m, type, code, param); 290} 291 292/* 293 * Generate an error packet of type error in response to bad IP6 packet. 294 */ 295void 296icmp6_error(struct mbuf *m, int type, int code, int param) 297{ 298 struct ip6_hdr *oip6, *nip6; 299 struct icmp6_hdr *icmp6; 300 u_int preplen; 301 int off; 302 int nxt; 303 304 ICMP6_STATINC(ICMP6_STAT_ERROR); 305 306 /* count per-type-code statistics */ 307 icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code); 308 309 if (m->m_flags & M_DECRYPTED) { 310 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 311 goto freeit; 312 } 313 314 if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) && 315 (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) 316 return; 317 oip6 = mtod(m, struct ip6_hdr *); 318 319 /* 320 * If the destination address of the erroneous packet is a multicast 321 * address, or the packet was sent using link-layer multicast, 322 * we should basically suppress sending an error (RFC 2463, Section 323 * 2.4). 324 * We have two exceptions (the item e.2 in that section): 325 * - the Pakcet Too Big message can be sent for path MTU discovery. 326 * - the Parameter Problem Message that can be allowed an icmp6 error 327 * in the option type field. This check has been done in 328 * ip6_unknown_opt(), so we can just check the type and code. 329 */ 330 if ((m->m_flags & (M_BCAST|M_MCAST) || 331 IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) && 332 (type != ICMP6_PACKET_TOO_BIG && 333 (type != ICMP6_PARAM_PROB || 334 code != ICMP6_PARAMPROB_OPTION))) 335 goto freeit; 336 337 /* 338 * RFC 2463, 2.4 (e.5): source address check. 339 * XXX: the case of anycast source? 340 */ 341 if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) || 342 IN6_IS_ADDR_MULTICAST(&oip6->ip6_src)) 343 goto freeit; 344 345 /* 346 * If we are about to send ICMPv6 against ICMPv6 error/redirect, 347 * don't do it. 348 */ 349 nxt = -1; 350 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt); 351 if (off >= 0 && nxt == IPPROTO_ICMPV6) { 352 struct icmp6_hdr *icp; 353 354 IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off, 355 sizeof(*icp)); 356 if (icp == NULL) { 357 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 358 return; 359 } 360 if (icp->icmp6_type < ICMP6_ECHO_REQUEST || 361 icp->icmp6_type == ND_REDIRECT) { 362 /* 363 * ICMPv6 error 364 * Special case: for redirect (which is 365 * informational) we must not send icmp6 error. 366 */ 367 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 368 goto freeit; 369 } else { 370 /* ICMPv6 informational - send the error */ 371 } 372 } 373#if 0 /* controversial */ 374 else if (off >= 0 && nxt == IPPROTO_ESP) { 375 /* 376 * It could be ICMPv6 error inside ESP. Take a safer side, 377 * don't respond. 378 */ 379 ICMP6_STATINC(ICMP6_STAT_CANTERROR); 380 goto freeit; 381 } 382#endif 383 else { 384 /* non-ICMPv6 - send the error */ 385 } 386 387 oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */ 388 389 /* Finally, do rate limitation check. */ 390 if (icmp6_ratelimit(&oip6->ip6_src, type, code)) { 391 ICMP6_STATINC(ICMP6_STAT_TOOFREQ); 392 goto freeit; 393 } 394 395 /* 396 * OK, ICMP6 can be generated. 397 */ 398 399 if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN) 400 m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len); 401 402 preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 403 M_PREPEND(m, preplen, M_DONTWAIT); 404 if (m && M_UNWRITABLE(m, preplen)) 405 m = m_pullup(m, preplen); 406 if (m == NULL) { 407 nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__); 408 return; 409 } 410 411 nip6 = mtod(m, struct ip6_hdr *); 412 nip6->ip6_src = oip6->ip6_src; 413 nip6->ip6_dst = oip6->ip6_dst; 414 415 in6_clearscope(&oip6->ip6_src); 416 in6_clearscope(&oip6->ip6_dst); 417 418 icmp6 = (struct icmp6_hdr *)(nip6 + 1); 419 icmp6->icmp6_type = type; 420 icmp6->icmp6_code = code; 421 icmp6->icmp6_pptr = htonl((u_int32_t)param); 422 423 /* 424 * icmp6_reflect() is designed to be in the input path. 425 * icmp6_error() can be called from both input and output path, 426 * and if we are in output path rcvif could contain bogus value. 427 * clear m->m_pkthdr.rcvif for safety, we should have enough scope 428 * information in ip header (nip6). 429 */ 430 m_reset_rcvif(m); 431 432 ICMP6_STATINC(ICMP6_STAT_OUTHIST + type); 433 icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */ 434 435 return; 436 437 freeit: 438 /* 439 * If we can't tell whether or not we can generate ICMP6, free it. 440 */ 441 m_freem(m); 442} 443 444/* 445 * Process a received ICMP6 message. 446 */ 447int 448icmp6_input(struct mbuf **mp, int *offp, int proto) 449{ 450 struct mbuf *m = *mp, *n; 451 struct ip6_hdr *ip6, *nip6; 452 struct icmp6_hdr *icmp6, *nicmp6; 453 int off = *offp; 454 int icmp6len = m->m_pkthdr.len - *offp; 455 int code, sum, noff, i; 456 struct ifnet *rcvif; 457 struct psref psref; 458 459 rcvif = m_get_rcvif_psref(m, &psref); 460#define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4) 461 KASSERT(ICMP6_MAXLEN < MCLBYTES); 462 icmp6_ifstat_inc(rcvif, ifs6_in_msg); 463 464 /* 465 * Locate icmp6 structure in mbuf, and check 466 * that not corrupted and of at least minimum length 467 */ 468 469 if (icmp6len < sizeof(struct icmp6_hdr)) { 470 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 471 icmp6_ifstat_inc(rcvif, ifs6_in_error); 472 goto freeit; 473 } 474 475 i = off + sizeof(*icmp6); 476 if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) { 477 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 478#if 0 /* m is 0 here */ 479 icmp6_ifstat_inc(rcvif, ifs6_in_error); 480#endif 481 goto freeit; 482 } 483 ip6 = mtod(m, struct ip6_hdr *); 484 /* 485 * calculate the checksum 486 */ 487 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 488 if (icmp6 == NULL) { 489 m_put_rcvif_psref(rcvif, &psref); 490 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 491 /* m is invalid */ 492 /*icmp6_ifstat_inc(rcvif, ifs6_in_error);*/ 493 return IPPROTO_DONE; 494 } 495 KASSERT(IP6_HDR_ALIGNED_P(icmp6)); 496 code = icmp6->icmp6_code; 497 498 if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) { 499 nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n", 500 icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)); 501 ICMP6_STATINC(ICMP6_STAT_CHECKSUM); 502 icmp6_ifstat_inc(rcvif, ifs6_in_error); 503 goto freeit; 504 } 505 506#if defined(NFAITH) && 0 < NFAITH 507 if (faithprefix(&ip6->ip6_dst)) { 508 /* 509 * Deliver very specific ICMP6 type only. 510 * This is important to deliver TOOBIG. Otherwise PMTUD 511 * will not work. 512 */ 513 switch (icmp6->icmp6_type) { 514 case ICMP6_DST_UNREACH: 515 case ICMP6_PACKET_TOO_BIG: 516 case ICMP6_TIME_EXCEEDED: 517 break; 518 default: 519 goto freeit; 520 } 521 } 522#endif 523 524 ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type); 525 526 switch (icmp6->icmp6_type) { 527 case ICMP6_DST_UNREACH: 528 icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach); 529 switch (code) { 530 case ICMP6_DST_UNREACH_NOROUTE: 531 code = PRC_UNREACH_NET; 532 break; 533 case ICMP6_DST_UNREACH_ADMIN: 534 icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib); 535 code = PRC_UNREACH_PROTOCOL; /* is this a good code? */ 536 break; 537 case ICMP6_DST_UNREACH_ADDR: 538 code = PRC_HOSTDEAD; 539 break; 540#ifdef COMPAT_RFC1885 541 case ICMP6_DST_UNREACH_NOTNEIGHBOR: 542 code = PRC_UNREACH_SRCFAIL; 543 break; 544#else 545 case ICMP6_DST_UNREACH_BEYONDSCOPE: 546 /* I mean "source address was incorrect." */ 547 code = PRC_UNREACH_NET; 548 break; 549#endif 550 case ICMP6_DST_UNREACH_NOPORT: 551 code = PRC_UNREACH_PORT; 552 break; 553 default: 554 goto badcode; 555 } 556 goto deliver; 557 558 case ICMP6_PACKET_TOO_BIG: 559 icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig); 560 561 /* 562 * MTU is checked in icmp6_mtudisc. 563 */ 564 code = PRC_MSGSIZE; 565 566 /* 567 * Updating the path MTU will be done after examining 568 * intermediate extension headers. 569 */ 570 goto deliver; 571 572 case ICMP6_TIME_EXCEEDED: 573 icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed); 574 switch (code) { 575 case ICMP6_TIME_EXCEED_TRANSIT: 576 code = PRC_TIMXCEED_INTRANS; 577 break; 578 case ICMP6_TIME_EXCEED_REASSEMBLY: 579 code = PRC_TIMXCEED_REASS; 580 break; 581 default: 582 goto badcode; 583 } 584 goto deliver; 585 586 case ICMP6_PARAM_PROB: 587 icmp6_ifstat_inc(rcvif, ifs6_in_paramprob); 588 switch (code) { 589 case ICMP6_PARAMPROB_NEXTHEADER: 590 code = PRC_UNREACH_PROTOCOL; 591 break; 592 case ICMP6_PARAMPROB_HEADER: 593 case ICMP6_PARAMPROB_OPTION: 594 code = PRC_PARAMPROB; 595 break; 596 default: 597 goto badcode; 598 } 599 goto deliver; 600 601 case ICMP6_ECHO_REQUEST: 602 icmp6_ifstat_inc(rcvif, ifs6_in_echo); 603 if (code != 0) 604 goto badcode; 605 /* 606 * Copy mbuf to send to two data paths: userland socket(s), 607 * and to the querier (echo reply). 608 * m: a copy for socket, n: a copy for querier 609 * 610 * If the first mbuf is shared, or the first mbuf is too short, 611 * copy the first part of the data into a fresh mbuf. 612 * Otherwise, we will wrongly overwrite both copies. 613 */ 614 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 615 /* Give up local */ 616 n = m; 617 m = NULL; 618 } else if (M_READONLY(n) || 619 n->m_len < off + sizeof(struct icmp6_hdr)) { 620 struct mbuf *n0 = n; 621 622 /* 623 * Prepare an internal mbuf. m_pullup() doesn't 624 * always copy the length we specified. 625 */ 626 if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 627 /* Give up local */ 628 n = m; 629 m = NULL; 630 } 631 m_freem(n0); 632 } 633 IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off, 634 sizeof(*nicmp6)); 635 if (nicmp6 == NULL) 636 goto freeit; 637 nicmp6->icmp6_type = ICMP6_ECHO_REPLY; 638 nicmp6->icmp6_code = 0; 639 if (n) { 640 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 641 icmp6s[ICMP6_STAT_REFLECT]++; 642 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++; 643 ICMP6_STAT_PUTREF(); 644 icmp6_reflect(n, off); 645 } 646 if (!m) 647 goto freeit; 648 break; 649 650 case ICMP6_ECHO_REPLY: 651 icmp6_ifstat_inc(rcvif, ifs6_in_echoreply); 652 if (code != 0) 653 goto badcode; 654 break; 655 656 case MLD_LISTENER_QUERY: 657 case MLD_LISTENER_REPORT: 658 if (icmp6len < sizeof(struct mld_hdr)) 659 goto badlen; 660 if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */ 661 icmp6_ifstat_inc(rcvif, ifs6_in_mldquery); 662 else 663 icmp6_ifstat_inc(rcvif, ifs6_in_mldreport); 664 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 665 /* give up local */ 666 mld_input(m, off); 667 m = NULL; 668 goto freeit; 669 } 670 mld_input(n, off); 671 /* m stays. */ 672 break; 673 674 case MLD_LISTENER_DONE: 675 icmp6_ifstat_inc(rcvif, ifs6_in_mlddone); 676 if (icmp6len < sizeof(struct mld_hdr)) /* necessary? */ 677 goto badlen; 678 break; /* nothing to be done in kernel */ 679 680 case MLD_MTRACE_RESP: 681 case MLD_MTRACE: 682 /* XXX: these two are experimental. not officially defined. */ 683 /* XXX: per-interface statistics? */ 684 break; /* just pass it to applications */ 685 686 case ICMP6_WRUREQUEST: /* ICMP6_FQDN_QUERY */ 687 { 688 enum { WRU, FQDN } mode; 689 690 if (!icmp6_nodeinfo) 691 break; 692 693 if (icmp6len == sizeof(struct icmp6_hdr) + 4) 694 mode = WRU; 695 else if (icmp6len >= sizeof(struct icmp6_nodeinfo)) 696 mode = FQDN; 697 else 698 goto badlen; 699 700 if (mode == FQDN) { 701 n = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 702 if (n) 703 n = ni6_input(n, off); 704 /* XXX meaningless if n == NULL */ 705 noff = sizeof(struct ip6_hdr); 706 } else { 707 u_char *p; 708 int maxhlen; 709 710 if ((icmp6_nodeinfo & 5) != 5) 711 break; 712 713 if (code != 0) 714 goto badcode; 715 MGETHDR(n, M_DONTWAIT, m->m_type); 716 if (n && ICMP6_MAXLEN > MHLEN) { 717 MCLGET(n, M_DONTWAIT); 718 if ((n->m_flags & M_EXT) == 0) { 719 m_free(n); 720 n = NULL; 721 } 722 } 723 if (n == NULL) { 724 /* Give up remote */ 725 break; 726 } 727 m_reset_rcvif(n); 728 n->m_len = 0; 729 maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN; 730 if (maxhlen < 0) 731 break; 732 if (maxhlen > hostnamelen) 733 maxhlen = hostnamelen; 734 /* 735 * Copy IPv6 and ICMPv6 only. 736 */ 737 nip6 = mtod(n, struct ip6_hdr *); 738 bcopy(ip6, nip6, sizeof(struct ip6_hdr)); 739 nicmp6 = (struct icmp6_hdr *)(nip6 + 1); 740 bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr)); 741 p = (u_char *)(nicmp6 + 1); 742 memset(p, 0, 4); 743 bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */ 744 noff = sizeof(struct ip6_hdr); 745 M_COPY_PKTHDR(n, m); /* just for rcvif */ 746 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 747 sizeof(struct icmp6_hdr) + 4 + maxhlen; 748 nicmp6->icmp6_type = ICMP6_WRUREPLY; 749 nicmp6->icmp6_code = 0; 750 } 751#undef hostnamelen 752 if (n) { 753 uint64_t *icmp6s = ICMP6_STAT_GETREF(); 754 icmp6s[ICMP6_STAT_REFLECT]++; 755 icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++; 756 ICMP6_STAT_PUTREF(); 757 icmp6_reflect(n, noff); 758 } 759 break; 760 } 761 762 case ICMP6_WRUREPLY: 763 if (code != 0) 764 goto badcode; 765 break; 766 767 case ND_ROUTER_SOLICIT: 768 icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit); 769 if (code != 0) 770 goto badcode; 771 if (icmp6len < sizeof(struct nd_router_solicit)) 772 goto badlen; 773 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 774 /* give up local */ 775 nd6_rs_input(m, off, icmp6len); 776 m = NULL; 777 goto freeit; 778 } 779 nd6_rs_input(n, off, icmp6len); 780 /* m stays. */ 781 break; 782 783 case ND_ROUTER_ADVERT: 784 icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert); 785 if (code != 0) 786 goto badcode; 787 if (icmp6len < sizeof(struct nd_router_advert)) 788 goto badlen; 789 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 790 /* give up local */ 791 nd6_ra_input(m, off, icmp6len); 792 m = NULL; 793 goto freeit; 794 } 795 nd6_ra_input(n, off, icmp6len); 796 /* m stays. */ 797 break; 798 799 case ND_NEIGHBOR_SOLICIT: 800 icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit); 801 if (code != 0) 802 goto badcode; 803 if (icmp6len < sizeof(struct nd_neighbor_solicit)) 804 goto badlen; 805 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 806 /* give up local */ 807 nd6_ns_input(m, off, icmp6len); 808 m = NULL; 809 goto freeit; 810 } 811 nd6_ns_input(n, off, icmp6len); 812 /* m stays. */ 813 break; 814 815 case ND_NEIGHBOR_ADVERT: 816 icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert); 817 if (code != 0) 818 goto badcode; 819 if (icmp6len < sizeof(struct nd_neighbor_advert)) 820 goto badlen; 821 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 822 /* give up local */ 823 nd6_na_input(m, off, icmp6len); 824 m = NULL; 825 goto freeit; 826 } 827 nd6_na_input(n, off, icmp6len); 828 /* m stays. */ 829 break; 830 831 case ND_REDIRECT: 832 icmp6_ifstat_inc(rcvif, ifs6_in_redirect); 833 if (code != 0) 834 goto badcode; 835 if (icmp6len < sizeof(struct nd_redirect)) 836 goto badlen; 837 if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) { 838 /* give up local */ 839 icmp6_redirect_input(m, off); 840 m = NULL; 841 goto freeit; 842 } 843 icmp6_redirect_input(n, off); 844 /* m stays. */ 845 break; 846 847 case ICMP6_ROUTER_RENUMBERING: 848 if (code != ICMP6_ROUTER_RENUMBERING_COMMAND && 849 code != ICMP6_ROUTER_RENUMBERING_RESULT) 850 goto badcode; 851 if (icmp6len < sizeof(struct icmp6_router_renum)) 852 goto badlen; 853 break; 854 855 default: 856 nd6log(LOG_DEBUG, "unknown type %d(src=%s, dst=%s, ifid=%d)\n", 857 icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src), 858 ip6_sprintf(&ip6->ip6_dst), 859 rcvif ? rcvif->if_index : 0); 860 if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) { 861 /* ICMPv6 error: MUST deliver it by spec... */ 862 code = PRC_NCMDS; 863 /* deliver */ 864 } else { 865 /* ICMPv6 informational: MUST not deliver */ 866 break; 867 } 868 deliver: 869 if (icmp6_notify_error(m, off, icmp6len, code)) { 870 /* In this case, m should've been freed. */ 871 m_put_rcvif_psref(rcvif, &psref); 872 return (IPPROTO_DONE); 873 } 874 break; 875 876 badcode: 877 ICMP6_STATINC(ICMP6_STAT_BADCODE); 878 break; 879 880 badlen: 881 ICMP6_STATINC(ICMP6_STAT_BADLEN); 882 break; 883 } 884 m_put_rcvif_psref(rcvif, &psref); 885 886 /* deliver the packet to appropriate sockets */ 887 icmp6_rip6_input(&m, *offp); 888 889 return IPPROTO_DONE; 890 891 freeit: 892 m_put_rcvif_psref(rcvif, &psref); 893 m_freem(m); 894 return IPPROTO_DONE; 895} 896 897static int 898icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code) 899{ 900 struct icmp6_hdr *icmp6; 901 struct ip6_hdr *eip6; 902 u_int32_t notifymtu; 903 struct sockaddr_in6 icmp6src, icmp6dst; 904 905 if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) { 906 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 907 goto freeit; 908 } 909 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 910 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 911 if (icmp6 == NULL) { 912 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 913 return (-1); 914 } 915 eip6 = (struct ip6_hdr *)(icmp6 + 1); 916 917 /* Detect the upper level protocol */ 918 { 919 void (*ctlfunc)(int, struct sockaddr *, void *); 920 u_int8_t nxt = eip6->ip6_nxt; 921 int eoff = off + sizeof(struct icmp6_hdr) + 922 sizeof(struct ip6_hdr); 923 struct ip6ctlparam ip6cp; 924 struct in6_addr *finaldst = NULL; 925 int icmp6type = icmp6->icmp6_type; 926 struct ip6_frag *fh; 927 struct ip6_rthdr *rth; 928 struct ip6_rthdr0 *rth0; 929 int rthlen; 930 struct ifnet *rcvif; 931 int s; 932 933 while (1) { /* XXX: should avoid infinite loop explicitly? */ 934 struct ip6_ext *eh; 935 936 switch (nxt) { 937 case IPPROTO_HOPOPTS: 938 case IPPROTO_DSTOPTS: 939 case IPPROTO_AH: 940 IP6_EXTHDR_GET(eh, struct ip6_ext *, m, 941 eoff, sizeof(*eh)); 942 if (eh == NULL) { 943 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 944 return (-1); 945 } 946 947 if (nxt == IPPROTO_AH) 948 eoff += (eh->ip6e_len + 2) << 2; 949 else 950 eoff += (eh->ip6e_len + 1) << 3; 951 nxt = eh->ip6e_nxt; 952 break; 953 case IPPROTO_ROUTING: 954 /* 955 * When the erroneous packet contains a 956 * routing header, we should examine the 957 * header to determine the final destination. 958 * Otherwise, we can't properly update 959 * information that depends on the final 960 * destination (e.g. path MTU). 961 */ 962 IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m, 963 eoff, sizeof(*rth)); 964 if (rth == NULL) { 965 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 966 return (-1); 967 } 968 rthlen = (rth->ip6r_len + 1) << 3; 969 /* 970 * XXX: currently there is no 971 * officially defined type other 972 * than type-0. 973 * Note that if the segment left field 974 * is 0, all intermediate hops must 975 * have been passed. 976 */ 977 if (rth->ip6r_segleft && 978 rth->ip6r_type == IPV6_RTHDR_TYPE_0) { 979 int hops; 980 981 IP6_EXTHDR_GET(rth0, 982 struct ip6_rthdr0 *, m, 983 eoff, rthlen); 984 if (rth0 == NULL) { 985 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 986 return (-1); 987 } 988 /* just ignore a bogus header */ 989 if ((rth0->ip6r0_len % 2) == 0 && 990 (hops = rth0->ip6r0_len/2)) 991 finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1); 992 } 993 eoff += rthlen; 994 nxt = rth->ip6r_nxt; 995 break; 996 case IPPROTO_FRAGMENT: 997 IP6_EXTHDR_GET(fh, struct ip6_frag *, m, 998 eoff, sizeof(*fh)); 999 if (fh == NULL) { 1000 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 1001 return (-1); 1002 } 1003 /* 1004 * Data after a fragment header is meaningless 1005 * unless it is the first fragment, but 1006 * we'll go to the notify label for path MTU 1007 * discovery. 1008 */ 1009 if (fh->ip6f_offlg & IP6F_OFF_MASK) 1010 goto notify; 1011 1012 eoff += sizeof(struct ip6_frag); 1013 nxt = fh->ip6f_nxt; 1014 break; 1015 default: 1016 /* 1017 * This case includes ESP and the No Next 1018 * Header. In such cases going to the notify 1019 * label does not have any meaning 1020 * (i.e. ctlfunc will be NULL), but we go 1021 * anyway since we might have to update 1022 * path MTU information. 1023 */ 1024 goto notify; 1025 } 1026 } 1027 notify: 1028 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, 1029 sizeof(*icmp6) + sizeof(struct ip6_hdr)); 1030 if (icmp6 == NULL) { 1031 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 1032 return (-1); 1033 } 1034 1035 /* 1036 * retrieve parameters from the inner IPv6 header, and convert 1037 * them into sockaddr structures. 1038 * XXX: there is no guarantee that the source or destination 1039 * addresses of the inner packet are in the same scope zone as 1040 * the addresses of the icmp packet. But there is no other 1041 * way to determine the zone. 1042 */ 1043 eip6 = (struct ip6_hdr *)(icmp6 + 1); 1044 1045 rcvif = m_get_rcvif(m, &s); 1046 sockaddr_in6_init(&icmp6dst, 1047 (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0); 1048 if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) { 1049 m_put_rcvif(rcvif, &s); 1050 goto freeit; 1051 } 1052 sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0); 1053 if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) { 1054 m_put_rcvif(rcvif, &s); 1055 goto freeit; 1056 } 1057 m_put_rcvif(rcvif, &s); 1058 1059 icmp6src.sin6_flowinfo = 1060 (eip6->ip6_flow & IPV6_FLOWLABEL_MASK); 1061 1062 if (finaldst == NULL) 1063 finaldst = &eip6->ip6_dst; 1064 ip6cp.ip6c_m = m; 1065 ip6cp.ip6c_icmp6 = icmp6; 1066 ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1); 1067 ip6cp.ip6c_off = eoff; 1068 ip6cp.ip6c_finaldst = finaldst; 1069 ip6cp.ip6c_src = &icmp6src; 1070 ip6cp.ip6c_nxt = nxt; 1071 1072 if (icmp6type == ICMP6_PACKET_TOO_BIG) { 1073 notifymtu = ntohl(icmp6->icmp6_mtu); 1074 ip6cp.ip6c_cmdarg = (void *)¬ifymtu; 1075 } 1076 1077 ctlfunc = (void (*)(int, struct sockaddr *, void *)) 1078 (inet6sw[ip6_protox[nxt]].pr_ctlinput); 1079 if (ctlfunc) { 1080 (void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst, 1081 &ip6cp); 1082 } 1083 } 1084 return (0); 1085 1086 freeit: 1087 m_freem(m); 1088 return (-1); 1089} 1090 1091void 1092icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated) 1093{ 1094 unsigned long rtcount; 1095 struct icmp6_mtudisc_callback *mc; 1096 struct in6_addr *dst = ip6cp->ip6c_finaldst; 1097 struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6; 1098 struct mbuf *m = ip6cp->ip6c_m; /* will be necessary for scope issue */ 1099 u_int mtu = ntohl(icmp6->icmp6_mtu); 1100 struct rtentry *rt = NULL; 1101 struct sockaddr_in6 sin6; 1102 struct ifnet *rcvif; 1103 int s; 1104 1105 /* 1106 * The MTU should not be less than the minimal IPv6 MTU except for the 1107 * hack in ip6_output/ip6_setpmtu where we always include a frag header. 1108 * In that one case, the MTU might be less than 1280. 1109 */ 1110 if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) { 1111 /* is the mtu even sane? */ 1112 if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8) 1113 return; 1114 if (!validated) 1115 return; 1116 mtu = IPV6_MMTU - sizeof(struct ip6_frag); 1117 } 1118 1119 /* 1120 * allow non-validated cases if memory is plenty, to make traffic 1121 * from non-connected pcb happy. 1122 */ 1123 rtcount = rt_timer_count(icmp6_mtudisc_timeout_q); 1124 if (validated) { 1125 if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat) 1126 return; 1127 else if (0 <= icmp6_mtudisc_lowat && 1128 rtcount > icmp6_mtudisc_lowat) { 1129 /* 1130 * XXX nuke a victim, install the new one. 1131 */ 1132 } 1133 } else { 1134 if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat) 1135 return; 1136 } 1137 1138 memset(&sin6, 0, sizeof(sin6)); 1139 sin6.sin6_family = PF_INET6; 1140 sin6.sin6_len = sizeof(struct sockaddr_in6); 1141 sin6.sin6_addr = *dst; 1142 s = pserialize_read_enter(); 1143 rcvif = m_get_rcvif(m, &s); 1144 if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) { 1145 m_put_rcvif(rcvif, &s); 1146 return; 1147 } 1148 m_put_rcvif(rcvif, &s); 1149 1150 rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6); 1151 1152 if (rt && (rt->rt_flags & RTF_HOST) && 1153 !(rt->rt_rmx.rmx_locks & RTV_MTU) && 1154 (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) { 1155 if (mtu < IN6_LINKMTU(rt->rt_ifp)) { 1156 ICMP6_STATINC(ICMP6_STAT_PMTUCHG); 1157 rt->rt_rmx.rmx_mtu = mtu; 1158 } 1159 } 1160 if (rt) { 1161 rtfree(rt); 1162 } 1163 1164 /* 1165 * Notify protocols that the MTU for this destination 1166 * has changed. 1167 */ 1168 for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL; 1169 mc = LIST_NEXT(mc, mc_list)) 1170 (*mc->mc_func)(&sin6.sin6_addr); 1171} 1172 1173/* 1174 * Process a Node Information Query packet, based on 1175 * draft-ietf-ipngwg-icmp-name-lookups-07. 1176 * 1177 * Spec incompatibilities: 1178 * - IPv6 Subject address handling 1179 * - IPv4 Subject address handling support missing 1180 * - Proxy reply (answer even if it's not for me) 1181 * - joins NI group address at in6_ifattach() time only, does not cope 1182 * with hostname changes by sethostname(3) 1183 */ 1184static struct mbuf * 1185ni6_input(struct mbuf *m, int off) 1186{ 1187 struct icmp6_nodeinfo *ni6, *nni6; 1188 struct mbuf *n = NULL; 1189 u_int16_t qtype; 1190 int subjlen; 1191 int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1192 struct ni_reply_fqdn *fqdn; 1193 int addrs; /* for NI_QTYPE_NODEADDR */ 1194 struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */ 1195 struct sockaddr_in6 sin6; /* ip6_dst */ 1196 struct in6_addr in6_subj; /* subject address */ 1197 struct ip6_hdr *ip6; 1198 int oldfqdn = 0; /* if 1, return pascal string (03 draft) */ 1199 char *subj = NULL; 1200 int s; 1201 struct ifnet *rcvif; 1202 1203 ip6 = mtod(m, struct ip6_hdr *); 1204 IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6)); 1205 if (ni6 == NULL) { 1206 /* m is already reclaimed */ 1207 return NULL; 1208 } 1209 1210 /* 1211 * Validate IPv6 destination address. 1212 * 1213 * The Responder must discard the Query without further processing 1214 * unless it is one of the Responder's unicast or anycast addresses, or 1215 * a link-local scope multicast address which the Responder has joined. 1216 * [icmp-name-lookups-07, Section 4.] 1217 */ 1218 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 1219 /* XXX scopeid */ 1220 if (ifa_ifwithaddr((struct sockaddr *)&sin6)) 1221 ; /* unicast/anycast, fine */ 1222 else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr)) 1223 ; /* link-local multicast, fine */ 1224 else 1225 goto bad; 1226 1227 /* validate query Subject field. */ 1228 qtype = ntohs(ni6->ni_qtype); 1229 subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo); 1230 switch (qtype) { 1231 case NI_QTYPE_NOOP: 1232 case NI_QTYPE_SUPTYPES: 1233 /* 07 draft */ 1234 if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0) 1235 break; 1236 /* FALLTHROUGH */ 1237 case NI_QTYPE_FQDN: 1238 case NI_QTYPE_NODEADDR: 1239 case NI_QTYPE_IPV4ADDR: 1240 switch (ni6->ni_code) { 1241 case ICMP6_NI_SUBJ_IPV6: 1242#if ICMP6_NI_SUBJ_IPV6 != 0 1243 case 0: 1244#endif 1245 /* 1246 * backward compatibility - try to accept 03 draft 1247 * format, where no Subject is present. 1248 */ 1249 if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 && 1250 subjlen == 0) { 1251 oldfqdn++; 1252 break; 1253 } 1254#if ICMP6_NI_SUBJ_IPV6 != 0 1255 if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6) 1256 goto bad; 1257#endif 1258 1259 if (subjlen != sizeof(sin6.sin6_addr)) 1260 goto bad; 1261 1262 /* 1263 * Validate Subject address. 1264 * 1265 * Not sure what exactly "address belongs to the node" 1266 * means in the spec, is it just unicast, or what? 1267 * 1268 * At this moment we consider Subject address as 1269 * "belong to the node" if the Subject address equals 1270 * to the IPv6 destination address; validation for 1271 * IPv6 destination address should have done enough 1272 * check for us. 1273 * 1274 * We do not do proxy at this moment. 1275 */ 1276 /* m_pulldown instead of copy? */ 1277 m_copydata(m, off + sizeof(struct icmp6_nodeinfo), 1278 subjlen, (void *)&in6_subj); 1279 rcvif = m_get_rcvif(m, &s); 1280 if (in6_setscope(&in6_subj, rcvif, NULL)) { 1281 m_put_rcvif(rcvif, &s); 1282 goto bad; 1283 } 1284 m_put_rcvif(rcvif, &s); 1285 1286 subj = (char *)&in6_subj; 1287 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj)) 1288 break; 1289 1290 /* 1291 * XXX if we are to allow other cases, we should really 1292 * be careful about scope here. 1293 * basically, we should disallow queries toward IPv6 1294 * destination X with subject Y, if scope(X) > scope(Y). 1295 * if we allow scope(X) > scope(Y), it will result in 1296 * information leakage across scope boundary. 1297 */ 1298 goto bad; 1299 1300 case ICMP6_NI_SUBJ_FQDN: 1301 /* 1302 * Validate Subject name with gethostname(3). 1303 * 1304 * The behavior may need some debate, since: 1305 * - we are not sure if the node has FQDN as 1306 * hostname (returned by gethostname(3)). 1307 * - the code does wildcard match for truncated names. 1308 * however, we are not sure if we want to perform 1309 * wildcard match, if gethostname(3) side has 1310 * truncated hostname. 1311 */ 1312 n = ni6_nametodns(hostname, hostnamelen, 0); 1313 if (!n || n->m_next || n->m_len == 0) 1314 goto bad; 1315 IP6_EXTHDR_GET(subj, char *, m, 1316 off + sizeof(struct icmp6_nodeinfo), subjlen); 1317 if (subj == NULL) 1318 goto bad; 1319 if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *), 1320 n->m_len)) { 1321 goto bad; 1322 } 1323 m_freem(n); 1324 n = NULL; 1325 break; 1326 1327 case ICMP6_NI_SUBJ_IPV4: /* XXX: to be implemented? */ 1328 default: 1329 goto bad; 1330 } 1331 break; 1332 } 1333 1334 /* refuse based on configuration. XXX ICMP6_NI_REFUSED? */ 1335 switch (qtype) { 1336 case NI_QTYPE_FQDN: 1337 if ((icmp6_nodeinfo & 1) == 0) 1338 goto bad; 1339 break; 1340 case NI_QTYPE_NODEADDR: 1341 case NI_QTYPE_IPV4ADDR: 1342 if ((icmp6_nodeinfo & 2) == 0) 1343 goto bad; 1344 break; 1345 } 1346 1347 /* guess reply length */ 1348 switch (qtype) { 1349 case NI_QTYPE_NOOP: 1350 break; /* no reply data */ 1351 case NI_QTYPE_SUPTYPES: 1352 replylen += sizeof(u_int32_t); 1353 break; 1354 case NI_QTYPE_FQDN: 1355 /* XXX will append an mbuf */ 1356 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1357 break; 1358 case NI_QTYPE_NODEADDR: 1359 addrs = ni6_addrs(ni6, m, &ifp, subj); 1360 if ((replylen += addrs * (sizeof(struct in6_addr) + 1361 sizeof(u_int32_t))) > MCLBYTES) 1362 replylen = MCLBYTES; /* XXX: will truncate pkt later */ 1363 break; 1364 case NI_QTYPE_IPV4ADDR: 1365 /* unsupported - should respond with unknown Qtype? */ 1366 goto bad; 1367 default: 1368 /* 1369 * XXX: We must return a reply with the ICMP6 code 1370 * `unknown Qtype' in this case. However we regard the case 1371 * as an FQDN query for backward compatibility. 1372 * Older versions set a random value to this field, 1373 * so it rarely varies in the defined qtypes. 1374 * But the mechanism is not reliable... 1375 * maybe we should obsolete older versions. 1376 */ 1377 qtype = NI_QTYPE_FQDN; 1378 /* XXX will append an mbuf */ 1379 replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen); 1380 oldfqdn++; 1381 break; 1382 } 1383 1384 /* allocate an mbuf to reply. */ 1385 MGETHDR(n, M_DONTWAIT, m->m_type); 1386 if (n == NULL) { 1387 m_freem(m); 1388 return (NULL); 1389 } 1390 M_MOVE_PKTHDR(n, m); /* just for rcvif */ 1391 if (replylen > MHLEN) { 1392 if (replylen > MCLBYTES) { 1393 /* 1394 * XXX: should we try to allocate more? But MCLBYTES 1395 * is probably much larger than IPV6_MMTU... 1396 */ 1397 goto bad; 1398 } 1399 MCLGET(n, M_DONTWAIT); 1400 if ((n->m_flags & M_EXT) == 0) { 1401 goto bad; 1402 } 1403 } 1404 n->m_pkthdr.len = n->m_len = replylen; 1405 1406 /* copy mbuf header and IPv6 + Node Information base headers */ 1407 bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr)); 1408 nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1); 1409 bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo)); 1410 1411 /* qtype dependent procedure */ 1412 switch (qtype) { 1413 case NI_QTYPE_NOOP: 1414 nni6->ni_code = ICMP6_NI_SUCCESS; 1415 nni6->ni_flags = 0; 1416 break; 1417 case NI_QTYPE_SUPTYPES: 1418 { 1419 u_int32_t v; 1420 nni6->ni_code = ICMP6_NI_SUCCESS; 1421 nni6->ni_flags = htons(0x0000); /* raw bitmap */ 1422 /* supports NOOP, SUPTYPES, FQDN, and NODEADDR */ 1423 v = (u_int32_t)htonl(0x0000000f); 1424 bcopy(&v, nni6 + 1, sizeof(u_int32_t)); 1425 break; 1426 } 1427 case NI_QTYPE_FQDN: 1428 nni6->ni_code = ICMP6_NI_SUCCESS; 1429 fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) + 1430 sizeof(struct ip6_hdr) + 1431 sizeof(struct icmp6_nodeinfo)); 1432 nni6->ni_flags = 0; /* XXX: meaningless TTL */ 1433 fqdn->ni_fqdn_ttl = 0; /* ditto. */ 1434 /* 1435 * XXX do we really have FQDN in variable "hostname"? 1436 */ 1437 n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn); 1438 if (n->m_next == NULL) 1439 goto bad; 1440 /* XXX we assume that n->m_next is not a chain */ 1441 if (n->m_next->m_next != NULL) 1442 goto bad; 1443 n->m_pkthdr.len += n->m_next->m_len; 1444 break; 1445 case NI_QTYPE_NODEADDR: 1446 { 1447 int lenlim, copied; 1448 1449 nni6->ni_code = ICMP6_NI_SUCCESS; 1450 n->m_pkthdr.len = n->m_len = 1451 sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo); 1452 lenlim = M_TRAILINGSPACE(n); 1453 copied = ni6_store_addrs(ni6, nni6, ifp, lenlim); 1454 /* XXX: reset mbuf length */ 1455 n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) + 1456 sizeof(struct icmp6_nodeinfo) + copied; 1457 break; 1458 } 1459 default: 1460 break; /* XXX impossible! */ 1461 } 1462 1463 nni6->ni_type = ICMP6_NI_REPLY; 1464 m_freem(m); 1465 return (n); 1466 1467 bad: 1468 m_freem(m); 1469 if (n) 1470 m_freem(n); 1471 return (NULL); 1472} 1473#undef hostnamelen 1474 1475#define isupper(x) ('A' <= (x) && (x) <= 'Z') 1476#define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z')) 1477#define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9')) 1478#define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x)) 1479 1480/* 1481 * make a mbuf with DNS-encoded string. no compression support. 1482 * 1483 * XXX names with less than 2 dots (like "foo" or "foo.section") will be 1484 * treated as truncated name (two \0 at the end). this is a wild guess. 1485 * 1486 * old - return pascal string if non-zero 1487 */ 1488static struct mbuf * 1489ni6_nametodns(const char *name, int namelen, int old) 1490{ 1491 struct mbuf *m; 1492 char *cp, *ep; 1493 const char *p, *q; 1494 int i, len, nterm; 1495 1496 if (old) 1497 len = namelen + 1; 1498 else 1499 len = MCLBYTES; 1500 1501 /* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */ 1502 MGET(m, M_DONTWAIT, MT_DATA); 1503 if (m && len > MLEN) { 1504 MCLGET(m, M_DONTWAIT); 1505 if ((m->m_flags & M_EXT) == 0) 1506 goto fail; 1507 } 1508 if (!m) 1509 goto fail; 1510 m->m_next = NULL; 1511 1512 if (old) { 1513 m->m_len = len; 1514 *mtod(m, char *) = namelen; 1515 bcopy(name, mtod(m, char *) + 1, namelen); 1516 return m; 1517 } else { 1518 m->m_len = 0; 1519 cp = mtod(m, char *); 1520 ep = mtod(m, char *) + M_TRAILINGSPACE(m); 1521 1522 /* if not certain about my name, return empty buffer */ 1523 if (namelen == 0) 1524 return m; 1525 1526 /* 1527 * guess if it looks like shortened hostname, or FQDN. 1528 * shortened hostname needs two trailing "\0". 1529 */ 1530 i = 0; 1531 for (p = name; p < name + namelen; p++) { 1532 if (*p && *p == '.') 1533 i++; 1534 } 1535 if (i < 2) 1536 nterm = 2; 1537 else 1538 nterm = 1; 1539 1540 p = name; 1541 while (cp < ep && p < name + namelen) { 1542 i = 0; 1543 for (q = p; q < name + namelen && *q && *q != '.'; q++) 1544 i++; 1545 /* result does not fit into mbuf */ 1546 if (cp + i + 1 >= ep) 1547 goto fail; 1548 /* 1549 * DNS label length restriction, RFC1035 page 8. 1550 * "i == 0" case is included here to avoid returning 1551 * 0-length label on "foo..bar". 1552 */ 1553 if (i <= 0 || i >= 64) 1554 goto fail; 1555 *cp++ = i; 1556 if (!isalpha(p[0]) || !isalnum(p[i - 1])) 1557 goto fail; 1558 while (i > 0) { 1559 if (!isalnum(*p) && *p != '-') 1560 goto fail; 1561 if (isupper(*p)) { 1562 *cp++ = tolower(*p); 1563 p++; 1564 } else 1565 *cp++ = *p++; 1566 i--; 1567 } 1568 p = q; 1569 if (p < name + namelen && *p == '.') 1570 p++; 1571 } 1572 /* termination */ 1573 if (cp + nterm >= ep) 1574 goto fail; 1575 while (nterm-- > 0) 1576 *cp++ = '\0'; 1577 m->m_len = cp - mtod(m, char *); 1578 return m; 1579 } 1580 1581 panic("should not reach here"); 1582 /* NOTREACHED */ 1583 1584 fail: 1585 if (m) 1586 m_freem(m); 1587 return NULL; 1588} 1589 1590/* 1591 * check if two DNS-encoded string matches. takes care of truncated 1592 * form (with \0\0 at the end). no compression support. 1593 * XXX upper/lowercase match (see RFC2065) 1594 */ 1595static int 1596ni6_dnsmatch(const char *a, int alen, const char *b, int blen) 1597{ 1598 const char *a0, *b0; 1599 int l; 1600 1601 /* simplest case - need validation? */ 1602 if (alen == blen && memcmp(a, b, alen) == 0) 1603 return 1; 1604 1605 a0 = a; 1606 b0 = b; 1607 1608 /* termination is mandatory */ 1609 if (alen < 2 || blen < 2) 1610 return 0; 1611 if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0') 1612 return 0; 1613 alen--; 1614 blen--; 1615 1616 while (a - a0 < alen && b - b0 < blen) { 1617 if (a - a0 + 1 > alen || b - b0 + 1 > blen) 1618 return 0; 1619 1620 if ((signed char)a[0] < 0 || (signed char)b[0] < 0) 1621 return 0; 1622 /* we don't support compression yet */ 1623 if (a[0] >= 64 || b[0] >= 64) 1624 return 0; 1625 1626 /* truncated case */ 1627 if (a[0] == 0 && a - a0 == alen - 1) 1628 return 1; 1629 if (b[0] == 0 && b - b0 == blen - 1) 1630 return 1; 1631 if (a[0] == 0 || b[0] == 0) 1632 return 0; 1633 1634 if (a[0] != b[0]) 1635 return 0; 1636 l = a[0]; 1637 if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen) 1638 return 0; 1639 if (memcmp(a + 1, b + 1, l) != 0) 1640 return 0; 1641 1642 a += 1 + l; 1643 b += 1 + l; 1644 } 1645 1646 if (a - a0 == alen && b - b0 == blen) 1647 return 1; 1648 else 1649 return 0; 1650} 1651 1652/* 1653 * calculate the number of addresses to be returned in the node info reply. 1654 */ 1655static int 1656ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, 1657 struct ifnet **ifpp, char *subj) 1658{ 1659 struct ifnet *ifp; 1660 struct in6_ifaddr *ifa6; 1661 struct ifaddr *ifa; 1662 struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */ 1663 int addrs = 0, addrsofif, iffound = 0; 1664 int niflags = ni6->ni_flags; 1665 1666 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) { 1667 switch (ni6->ni_code) { 1668 case ICMP6_NI_SUBJ_IPV6: 1669 if (subj == NULL) /* must be impossible... */ 1670 return (0); 1671 subj_ip6 = (struct sockaddr_in6 *)subj; 1672 break; 1673 default: 1674 /* 1675 * XXX: we only support IPv6 subject address for 1676 * this Qtype. 1677 */ 1678 return (0); 1679 } 1680 } 1681 1682 IFNET_READER_FOREACH(ifp) { 1683 addrsofif = 0; 1684 IFADDR_FOREACH(ifa, ifp) { 1685 if (ifa->ifa_addr->sa_family != AF_INET6) 1686 continue; 1687 ifa6 = (struct in6_ifaddr *)ifa; 1688 1689 if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 && 1690 IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr, 1691 &ifa6->ia_addr.sin6_addr)) 1692 iffound = 1; 1693 1694 /* 1695 * IPv4-mapped addresses can only be returned by a 1696 * Node Information proxy, since they represent 1697 * addresses of IPv4-only nodes, which perforce do 1698 * not implement this protocol. 1699 * [icmp-name-lookups-07, Section 5.4] 1700 * So we don't support NI_NODEADDR_FLAG_COMPAT in 1701 * this function at this moment. 1702 */ 1703 1704 /* What do we have to do about ::1? */ 1705 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1706 case IPV6_ADDR_SCOPE_LINKLOCAL: 1707 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1708 continue; 1709 break; 1710 case IPV6_ADDR_SCOPE_SITELOCAL: 1711 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1712 continue; 1713 break; 1714 case IPV6_ADDR_SCOPE_GLOBAL: 1715 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1716 continue; 1717 break; 1718 default: 1719 continue; 1720 } 1721 1722 /* 1723 * check if anycast is okay. 1724 * XXX: just experimental. not in the spec. 1725 */ 1726 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1727 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1728 continue; /* we need only unicast addresses */ 1729 1730 addrsofif++; /* count the address */ 1731 } 1732 if (iffound) { 1733 *ifpp = ifp; 1734 return (addrsofif); 1735 } 1736 1737 addrs += addrsofif; 1738 } 1739 1740 return (addrs); 1741} 1742 1743static int 1744ni6_store_addrs(struct icmp6_nodeinfo *ni6, 1745 struct icmp6_nodeinfo *nni6, struct ifnet *ifp0, 1746 int resid) 1747{ 1748 struct ifnet *ifp = ifp0 ? ifp0 : IFNET_READER_FIRST(); 1749 struct in6_ifaddr *ifa6; 1750 struct ifaddr *ifa; 1751 struct ifnet *ifp_dep = NULL; 1752 int copied = 0, allow_deprecated = 0; 1753 u_char *cp = (u_char *)(nni6 + 1); 1754 int niflags = ni6->ni_flags; 1755 u_int32_t ltime; 1756 1757 if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL)) 1758 return (0); /* needless to copy */ 1759 1760 again: 1761 1762 for (; ifp; ifp = IFNET_READER_NEXT(ifp)) 1763 { 1764 IFADDR_FOREACH(ifa, ifp) { 1765 if (ifa->ifa_addr->sa_family != AF_INET6) 1766 continue; 1767 ifa6 = (struct in6_ifaddr *)ifa; 1768 1769 if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 && 1770 allow_deprecated == 0) { 1771 /* 1772 * prefererred address should be put before 1773 * deprecated addresses. 1774 */ 1775 1776 /* record the interface for later search */ 1777 if (ifp_dep == NULL) 1778 ifp_dep = ifp; 1779 1780 continue; 1781 } 1782 else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 && 1783 allow_deprecated != 0) 1784 continue; /* we now collect deprecated addrs */ 1785 1786 /* What do we have to do about ::1? */ 1787 switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) { 1788 case IPV6_ADDR_SCOPE_LINKLOCAL: 1789 if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0) 1790 continue; 1791 break; 1792 case IPV6_ADDR_SCOPE_SITELOCAL: 1793 if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0) 1794 continue; 1795 break; 1796 case IPV6_ADDR_SCOPE_GLOBAL: 1797 if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0) 1798 continue; 1799 break; 1800 default: 1801 continue; 1802 } 1803 1804 /* 1805 * check if anycast is okay. 1806 * XXX: just experimental. not in the spec. 1807 */ 1808 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 && 1809 (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0) 1810 continue; 1811 1812 /* now we can copy the address */ 1813 if (resid < sizeof(struct in6_addr) + 1814 sizeof(u_int32_t)) { 1815 /* 1816 * We give up much more copy. 1817 * Set the truncate flag and return. 1818 */ 1819 nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE; 1820 return (copied); 1821 } 1822 1823 /* 1824 * Set the TTL of the address. 1825 * The TTL value should be one of the following 1826 * according to the specification: 1827 * 1828 * 1. The remaining lifetime of a DHCP lease on the 1829 * address, or 1830 * 2. The remaining Valid Lifetime of a prefix from 1831 * which the address was derived through Stateless 1832 * Autoconfiguration. 1833 * 1834 * Note that we currently do not support stateful 1835 * address configuration by DHCPv6, so the former 1836 * case can't happen. 1837 * 1838 * TTL must be 2^31 > TTL >= 0. 1839 */ 1840 if (ifa6->ia6_lifetime.ia6t_expire == 0) 1841 ltime = ND6_INFINITE_LIFETIME; 1842 else { 1843 if (ifa6->ia6_lifetime.ia6t_expire > 1844 time_uptime) 1845 ltime = ifa6->ia6_lifetime.ia6t_expire - 1846 time_uptime; 1847 else 1848 ltime = 0; 1849 } 1850 if (ltime > 0x7fffffff) 1851 ltime = 0x7fffffff; 1852 ltime = htonl(ltime); 1853 1854 bcopy(<ime, cp, sizeof(u_int32_t)); 1855 cp += sizeof(u_int32_t); 1856 1857 /* copy the address itself */ 1858 bcopy(&ifa6->ia_addr.sin6_addr, cp, 1859 sizeof(struct in6_addr)); 1860 in6_clearscope((struct in6_addr *)cp); /* XXX */ 1861 cp += sizeof(struct in6_addr); 1862 1863 resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1864 copied += (sizeof(struct in6_addr) + sizeof(u_int32_t)); 1865 } 1866 if (ifp0) /* we need search only on the specified IF */ 1867 break; 1868 } 1869 1870 if (allow_deprecated == 0 && ifp_dep != NULL) { 1871 ifp = ifp_dep; 1872 allow_deprecated = 1; 1873 1874 goto again; 1875 } 1876 1877 return (copied); 1878} 1879 1880/* 1881 * XXX almost dup'ed code with rip6_input. 1882 */ 1883static int 1884icmp6_rip6_input(struct mbuf **mp, int off) 1885{ 1886 struct mbuf *m = *mp; 1887 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 1888 struct inpcb_hdr *inph; 1889 struct in6pcb *in6p; 1890 struct in6pcb *last = NULL; 1891 struct sockaddr_in6 rip6src; 1892 struct icmp6_hdr *icmp6; 1893 struct mbuf *opts = NULL; 1894 1895 IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6)); 1896 if (icmp6 == NULL) { 1897 /* m is already reclaimed */ 1898 return IPPROTO_DONE; 1899 } 1900 1901 /* 1902 * XXX: the address may have embedded scope zone ID, which should be 1903 * hidden from applications. 1904 */ 1905 sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0); 1906 if (sa6_recoverscope(&rip6src)) { 1907 m_freem(m); 1908 return (IPPROTO_DONE); 1909 } 1910 1911 TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) { 1912 in6p = (struct in6pcb *)inph; 1913 if (in6p->in6p_af != AF_INET6) 1914 continue; 1915 if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6) 1916 continue; 1917 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) && 1918 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst)) 1919 continue; 1920 if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) && 1921 !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src)) 1922 continue; 1923 if (in6p->in6p_icmp6filt 1924 && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type, 1925 in6p->in6p_icmp6filt)) 1926 continue; 1927 if (last) { 1928 struct mbuf *n; 1929 if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) { 1930 if (last->in6p_flags & IN6P_CONTROLOPTS) 1931 ip6_savecontrol(last, &opts, ip6, n); 1932 /* strip intermediate headers */ 1933 m_adj(n, off); 1934 if (sbappendaddr(&last->in6p_socket->so_rcv, 1935 (struct sockaddr *)&rip6src, 1936 n, opts) == 0) { 1937 /* should notify about lost packet */ 1938 m_freem(n); 1939 if (opts) 1940 m_freem(opts); 1941 } else 1942 sorwakeup(last->in6p_socket); 1943 opts = NULL; 1944 } 1945 } 1946 last = in6p; 1947 } 1948 if (last) { 1949 if (last->in6p_flags & IN6P_CONTROLOPTS) 1950 ip6_savecontrol(last, &opts, ip6, m); 1951 /* strip intermediate headers */ 1952 m_adj(m, off); 1953 if (sbappendaddr(&last->in6p_socket->so_rcv, 1954 (struct sockaddr *)&rip6src, m, opts) == 0) { 1955 m_freem(m); 1956 if (opts) 1957 m_freem(opts); 1958 } else 1959 sorwakeup(last->in6p_socket); 1960 } else { 1961 m_freem(m); 1962 IP6_STATDEC(IP6_STAT_DELIVERED); 1963 } 1964 return IPPROTO_DONE; 1965} 1966 1967/* 1968 * Reflect the ip6 packet back to the source. 1969 * OFF points to the icmp6 header, counted from the top of the mbuf. 1970 * 1971 * Note: RFC 1885 required that an echo reply should be truncated if it 1972 * did not fit in with (return) path MTU, and KAME code supported the 1973 * behavior. However, as a clarification after the RFC, this limitation 1974 * was removed in a revised version of the spec, RFC 2463. We had kept the 1975 * old behavior, with a (non-default) ifdef block, while the new version of 1976 * the spec was an internet-draft status, and even after the new RFC was 1977 * published. But it would rather make sense to clean the obsoleted part 1978 * up, and to make the code simpler at this stage. 1979 */ 1980void 1981icmp6_reflect(struct mbuf *m, size_t off) 1982{ 1983 struct ip6_hdr *ip6; 1984 struct icmp6_hdr *icmp6; 1985 const struct in6_ifaddr *ia; 1986 const struct ip6aux *ip6a; 1987 int plen; 1988 int type, code; 1989 struct ifnet *outif = NULL; 1990 struct in6_addr origdst; 1991 const struct in6_addr *src = NULL; 1992 struct ifnet *rcvif; 1993 int s; 1994 1995 /* too short to reflect */ 1996 if (off < sizeof(struct ip6_hdr)) { 1997 nd6log(LOG_DEBUG, 1998 "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n", 1999 (u_long)off, (u_long)sizeof(struct ip6_hdr), 2000 __FILE__, __LINE__); 2001 goto bad; 2002 } 2003 2004 /* 2005 * If there are extra headers between IPv6 and ICMPv6, strip 2006 * off that header first. 2007 */ 2008#ifdef DIAGNOSTIC 2009 if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN) 2010 panic("assumption failed in icmp6_reflect"); 2011#endif 2012 if (off > sizeof(struct ip6_hdr)) { 2013 size_t l; 2014 struct ip6_hdr nip6; 2015 2016 l = off - sizeof(struct ip6_hdr); 2017 m_copydata(m, 0, sizeof(nip6), (void *)&nip6); 2018 m_adj(m, l); 2019 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2020 if (m->m_len < l) { 2021 if ((m = m_pullup(m, l)) == NULL) 2022 return; 2023 } 2024 bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6)); 2025 } else /* off == sizeof(struct ip6_hdr) */ { 2026 size_t l; 2027 l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr); 2028 if (m->m_len < l) { 2029 if ((m = m_pullup(m, l)) == NULL) 2030 return; 2031 } 2032 } 2033 plen = m->m_pkthdr.len - sizeof(struct ip6_hdr); 2034 ip6 = mtod(m, struct ip6_hdr *); 2035 ip6->ip6_nxt = IPPROTO_ICMPV6; 2036 icmp6 = (struct icmp6_hdr *)(ip6 + 1); 2037 type = icmp6->icmp6_type; /* keep type for statistics */ 2038 code = icmp6->icmp6_code; /* ditto. */ 2039 2040 origdst = ip6->ip6_dst; 2041 /* 2042 * ip6_input() drops a packet if its src is multicast. 2043 * So, the src is never multicast. 2044 */ 2045 ip6->ip6_dst = ip6->ip6_src; 2046 2047 /* 2048 * If the incoming packet was addressed directly to us (i.e. unicast), 2049 * use dst as the src for the reply. 2050 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible 2051 * (for example) when we encounter an error while forwarding procedure 2052 * destined to a duplicated address of ours. 2053 * Note that ip6_getdstifaddr() may fail if we are in an error handling 2054 * procedure of an outgoing packet of our own, in which case we need 2055 * to search in the ifaddr list. 2056 */ 2057 if (IN6_IS_ADDR_MULTICAST(&origdst)) 2058 ; 2059 else if ((ip6a = ip6_getdstifaddr(m)) != NULL) { 2060 if ((ip6a->ip6a_flags & 2061 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2062 src = &ip6a->ip6a_src; 2063 } else { 2064 union { 2065 struct sockaddr_in6 sin6; 2066 struct sockaddr sa; 2067 } u; 2068 2069 sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0); 2070 2071 ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa); 2072 2073 if (ia == NULL) 2074 ; 2075 else if ((ia->ia6_flags & 2076 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) 2077 src = &ia->ia_addr.sin6_addr; 2078 } 2079 2080 if (src == NULL) { 2081 int e; 2082 struct sockaddr_in6 sin6; 2083 struct route ro; 2084 2085 /* 2086 * This case matches to multicasts, our anycast, or unicasts 2087 * that we do not own. Select a source address based on the 2088 * source address of the erroneous packet. 2089 */ 2090 /* zone ID should be embedded */ 2091 sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0); 2092 2093 memset(&ro, 0, sizeof(ro)); 2094 src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL, &e); 2095 rtcache_free(&ro); 2096 if (src == NULL) { 2097 nd6log(LOG_DEBUG, 2098 "source can't be determined: " 2099 "dst=%s, error=%d\n", 2100 ip6_sprintf(&sin6.sin6_addr), e); 2101 goto bad; 2102 } 2103 } 2104 2105 ip6->ip6_src = *src; 2106 ip6->ip6_flow = 0; 2107 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2108 ip6->ip6_vfc |= IPV6_VERSION; 2109 ip6->ip6_nxt = IPPROTO_ICMPV6; 2110 rcvif = m_get_rcvif(m, &s); 2111 if (rcvif) { 2112 /* XXX: This may not be the outgoing interface */ 2113 ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim; 2114 } else 2115 ip6->ip6_hlim = ip6_defhlim; 2116 m_put_rcvif(rcvif, &s); 2117 2118 m->m_pkthdr.csum_flags = 0; 2119 icmp6->icmp6_cksum = 0; 2120 icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6, 2121 sizeof(struct ip6_hdr), plen); 2122 2123 /* 2124 * XXX option handling 2125 */ 2126 2127 m->m_flags &= ~(M_BCAST|M_MCAST); 2128 2129 /* 2130 * To avoid a "too big" situation at an intermediate router 2131 * and the path MTU discovery process, specify the IPV6_MINMTU flag. 2132 * Note that only echo and node information replies are affected, 2133 * since the length of ICMP6 errors is limited to the minimum MTU. 2134 */ 2135 if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) 2136 != 0 && outif) 2137 icmp6_ifstat_inc(outif, ifs6_out_error); 2138 if (outif) 2139 icmp6_ifoutstat_inc(outif, type, code); 2140 2141 return; 2142 2143 bad: 2144 m_freem(m); 2145 return; 2146} 2147 2148static const char * 2149icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6, 2150 struct in6_addr *tgt6) 2151{ 2152 static char buf[1024]; 2153 snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)", 2154 ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6)); 2155 return buf; 2156} 2157 2158void 2159icmp6_redirect_input(struct mbuf *m, int off) 2160{ 2161 struct ifnet *ifp; 2162 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 2163 struct nd_redirect *nd_rd; 2164 int icmp6len = ntohs(ip6->ip6_plen); 2165 char *lladdr = NULL; 2166 int lladdrlen = 0; 2167 struct rtentry *rt = NULL; 2168 int is_router; 2169 int is_onlink; 2170 struct in6_addr src6 = ip6->ip6_src; 2171 struct in6_addr redtgt6; 2172 struct in6_addr reddst6; 2173 union nd_opts ndopts; 2174 struct psref psref; 2175 2176 ifp = m_get_rcvif_psref(m, &psref); 2177 if (ifp == NULL) 2178 goto freeit; 2179 2180 /* XXX if we are router, we don't update route by icmp6 redirect */ 2181 if (ip6_forwarding) 2182 goto freeit; 2183 if (!icmp6_rediraccept) 2184 goto freeit; 2185 2186 IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len); 2187 if (nd_rd == NULL) { 2188 ICMP6_STATINC(ICMP6_STAT_TOOSHORT); 2189 m_put_rcvif_psref(ifp, &psref); 2190 return; 2191 } 2192 redtgt6 = nd_rd->nd_rd_target; 2193 reddst6 = nd_rd->nd_rd_dst; 2194 2195 if (in6_setscope(&redtgt6, ifp, NULL) || 2196 in6_setscope(&reddst6, ifp, NULL)) { 2197 goto freeit; 2198 } 2199 2200 /* validation */ 2201 if (!IN6_IS_ADDR_LINKLOCAL(&src6)) { 2202 nd6log(LOG_ERR, 2203 "ICMP6 redirect sent from %s rejected; " 2204 "must be from linklocal\n", ip6_sprintf(&src6)); 2205 goto bad; 2206 } 2207 if (ip6->ip6_hlim != 255) { 2208 nd6log(LOG_ERR, 2209 "ICMP6 redirect sent from %s rejected; " 2210 "hlim=%d (must be 255)\n", 2211 ip6_sprintf(&src6), ip6->ip6_hlim); 2212 goto bad; 2213 } 2214 { 2215 /* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */ 2216 struct sockaddr_in6 sin6; 2217 struct in6_addr *gw6; 2218 2219 sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0); 2220 rt = rtalloc1((struct sockaddr *)&sin6, 0); 2221 if (rt) { 2222 if (rt->rt_gateway == NULL || 2223 rt->rt_gateway->sa_family != AF_INET6) { 2224 nd6log(LOG_ERR, 2225 "ICMP6 redirect rejected; no route " 2226 "with inet6 gateway found for redirect dst: %s\n", 2227 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2228 rtfree(rt); 2229 goto bad; 2230 } 2231 2232 gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr); 2233 if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) { 2234 nd6log(LOG_ERR, 2235 "ICMP6 redirect rejected; " 2236 "not equal to gw-for-src=%s (must be same): %s\n", 2237 ip6_sprintf(gw6), 2238 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2239 rtfree(rt); 2240 goto bad; 2241 } 2242 } else { 2243 nd6log(LOG_ERR, "ICMP6 redirect rejected; " 2244 "no route found for redirect dst: %s\n", 2245 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2246 goto bad; 2247 } 2248 rtfree(rt); 2249 rt = NULL; 2250 } 2251 if (IN6_IS_ADDR_MULTICAST(&reddst6)) { 2252 nd6log(LOG_ERR, "ICMP6 redirect rejected; " 2253 "redirect dst must be unicast: %s\n", 2254 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2255 goto bad; 2256 } 2257 2258 is_router = is_onlink = 0; 2259 if (IN6_IS_ADDR_LINKLOCAL(&redtgt6)) 2260 is_router = 1; /* router case */ 2261 if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0) 2262 is_onlink = 1; /* on-link destination case */ 2263 if (!is_router && !is_onlink) { 2264 nd6log(LOG_ERR, "ICMP6 redirect rejected; " 2265 "neither router case nor onlink case: %s\n", 2266 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2267 goto bad; 2268 } 2269 /* validation passed */ 2270 2271 icmp6len -= sizeof(*nd_rd); 2272 nd6_option_init(nd_rd + 1, icmp6len, &ndopts); 2273 if (nd6_options(&ndopts) < 0) { 2274 nd6log(LOG_INFO, "invalid ND option, rejected: %s\n", 2275 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2276 /* nd6_options have incremented stats */ 2277 goto freeit; 2278 } 2279 2280 if (ndopts.nd_opts_tgt_lladdr) { 2281 lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1); 2282 lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3; 2283 } 2284 2285 if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) { 2286 nd6log(LOG_INFO, "lladdrlen mismatch for %s " 2287 "(if %d, icmp6 packet %d): %s\n", 2288 ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2, 2289 icmp6_redirect_diag(&src6, &reddst6, &redtgt6)); 2290 goto bad; 2291 } 2292 2293 /* RFC 2461 8.3 */ 2294 nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT, 2295 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER); 2296 2297 m_put_rcvif_psref(ifp, &psref); 2298 ifp = NULL; 2299 2300 if (!is_onlink) { /* better router case. perform rtredirect. */ 2301 /* perform rtredirect */ 2302 struct sockaddr_in6 sdst; 2303 struct sockaddr_in6 sgw; 2304 struct sockaddr_in6 ssrc; 2305 unsigned long rtcount; 2306 struct rtentry *newrt = NULL; 2307 2308 /* 2309 * do not install redirect route, if the number of entries 2310 * is too much (> hiwat). note that, the node (= host) will 2311 * work just fine even if we do not install redirect route 2312 * (there will be additional hops, though). 2313 */ 2314 rtcount = rt_timer_count(icmp6_redirect_timeout_q); 2315 if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes) 2316 goto freeit; 2317 if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat) 2318 goto freeit; 2319 else if (0 <= icmp6_redirect_lowat && 2320 rtcount > icmp6_redirect_lowat) { 2321 /* 2322 * XXX nuke a victim, install the new one. 2323 */ 2324 } 2325 2326 memset(&sdst, 0, sizeof(sdst)); 2327 memset(&sgw, 0, sizeof(sgw)); 2328 memset(&ssrc, 0, sizeof(ssrc)); 2329 sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6; 2330 sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len = 2331 sizeof(struct sockaddr_in6); 2332 bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr)); 2333 bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr)); 2334 bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr)); 2335 rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw, 2336 NULL, RTF_GATEWAY | RTF_HOST, 2337 (struct sockaddr *)&ssrc, 2338 &newrt); 2339 2340 if (newrt) { 2341 (void)rt_timer_add(newrt, icmp6_redirect_timeout, 2342 icmp6_redirect_timeout_q); 2343 rtfree(newrt); 2344 } 2345 } 2346 /* finally update cached route in each socket via pfctlinput */ 2347 { 2348 struct sockaddr_in6 sdst; 2349 2350 sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0); 2351 pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst); 2352#if defined(IPSEC) 2353 if (ipsec_used) 2354 key_sa_routechange((struct sockaddr *)&sdst); 2355#endif 2356 } 2357 2358 freeit: 2359 if (ifp != NULL) 2360 m_put_rcvif_psref(ifp, &psref); 2361 m_freem(m); 2362 return; 2363 2364 bad: 2365 m_put_rcvif_psref(ifp, &psref); 2366 ICMP6_STATINC(ICMP6_STAT_BADREDIRECT); 2367 m_freem(m); 2368} 2369 2370void 2371icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) 2372{ 2373 struct ifnet *ifp; /* my outgoing interface */ 2374 struct in6_addr *ifp_ll6; 2375 struct in6_addr *nexthop; 2376 struct ip6_hdr *sip6; /* m0 as struct ip6_hdr */ 2377 struct mbuf *m = NULL; /* newly allocated one */ 2378 struct ip6_hdr *ip6; /* m as struct ip6_hdr */ 2379 struct nd_redirect *nd_rd; 2380 size_t maxlen; 2381 u_char *p; 2382 struct sockaddr_in6 src_sa; 2383 2384 icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0); 2385 2386 /* if we are not router, we don't send icmp6 redirect */ 2387 if (!ip6_forwarding) 2388 goto fail; 2389 2390 /* sanity check */ 2391 if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp)) 2392 goto fail; 2393 2394 /* 2395 * Address check: 2396 * the source address must identify a neighbor, and 2397 * the destination address must not be a multicast address 2398 * [RFC 2461, sec 8.2] 2399 */ 2400 sip6 = mtod(m0, struct ip6_hdr *); 2401 sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0); 2402 if (nd6_is_addr_neighbor(&src_sa, ifp) == 0) 2403 goto fail; 2404 if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst)) 2405 goto fail; /* what should we do here? */ 2406 2407 /* rate limit */ 2408 if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0)) 2409 goto fail; 2410 2411 /* 2412 * Since we are going to append up to 1280 bytes (= IPV6_MMTU), 2413 * we almost always ask for an mbuf cluster for simplicity. 2414 * (MHLEN < IPV6_MMTU is almost always true) 2415 */ 2416 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2417 if (m && IPV6_MMTU >= MHLEN) { 2418#if IPV6_MMTU >= MCLBYTES 2419 _MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT); 2420#else 2421 MCLGET(m, M_DONTWAIT); 2422#endif 2423 } 2424 2425 if (!m) 2426 goto fail; 2427 m_reset_rcvif(m); 2428 m->m_len = 0; 2429 maxlen = M_TRAILINGSPACE(m); 2430 maxlen = min(IPV6_MMTU, maxlen); 2431 /* just for safety */ 2432 if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) + 2433 ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) { 2434 goto fail; 2435 } 2436 2437 { 2438 /* get ip6 linklocal address for ifp(my outgoing interface). */ 2439 struct in6_ifaddr *ia; 2440 if ((ia = in6ifa_ifpforlinklocal(ifp, 2441 IN6_IFF_NOTREADY| 2442 IN6_IFF_ANYCAST)) == NULL) 2443 goto fail; 2444 ifp_ll6 = &ia->ia_addr.sin6_addr; 2445 } 2446 2447 /* get ip6 linklocal address for the router. */ 2448 if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) { 2449 struct sockaddr_in6 *sin6; 2450 sin6 = (struct sockaddr_in6 *)rt->rt_gateway; 2451 nexthop = &sin6->sin6_addr; 2452 if (!IN6_IS_ADDR_LINKLOCAL(nexthop)) 2453 nexthop = NULL; 2454 } else 2455 nexthop = NULL; 2456 2457 /* ip6 */ 2458 ip6 = mtod(m, struct ip6_hdr *); 2459 ip6->ip6_flow = 0; 2460 ip6->ip6_vfc &= ~IPV6_VERSION_MASK; 2461 ip6->ip6_vfc |= IPV6_VERSION; 2462 /* ip6->ip6_plen will be set later */ 2463 ip6->ip6_nxt = IPPROTO_ICMPV6; 2464 ip6->ip6_hlim = 255; 2465 /* ip6->ip6_src must be linklocal addr for my outgoing if. */ 2466 bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr)); 2467 bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr)); 2468 2469 /* ND Redirect */ 2470 nd_rd = (struct nd_redirect *)(ip6 + 1); 2471 nd_rd->nd_rd_type = ND_REDIRECT; 2472 nd_rd->nd_rd_code = 0; 2473 nd_rd->nd_rd_reserved = 0; 2474 if (rt->rt_flags & RTF_GATEWAY) { 2475 /* 2476 * nd_rd->nd_rd_target must be a link-local address in 2477 * better router cases. 2478 */ 2479 if (!nexthop) 2480 goto fail; 2481 bcopy(nexthop, &nd_rd->nd_rd_target, 2482 sizeof(nd_rd->nd_rd_target)); 2483 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2484 sizeof(nd_rd->nd_rd_dst)); 2485 } else { 2486 /* make sure redtgt == reddst */ 2487 nexthop = &sip6->ip6_dst; 2488 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target, 2489 sizeof(nd_rd->nd_rd_target)); 2490 bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst, 2491 sizeof(nd_rd->nd_rd_dst)); 2492 } 2493 2494 p = (u_char *)(nd_rd + 1); 2495 2496 { 2497 /* target lladdr option */ 2498 struct llentry *ln = NULL; 2499 int len; 2500 struct nd_opt_hdr *nd_opt; 2501 char *lladdr; 2502 2503 ln = nd6_lookup(nexthop, ifp, false); 2504 if (ln == NULL) 2505 goto nolladdropt; 2506 len = sizeof(*nd_opt) + ifp->if_addrlen; 2507 len = (len + 7) & ~7; /* round by 8 */ 2508 /* safety check */ 2509 if (len + (p - (u_char *)ip6) > maxlen) { 2510 LLE_RUNLOCK(ln); 2511 goto nolladdropt; 2512 } 2513 if (ln->la_flags & LLE_VALID) { 2514 nd_opt = (struct nd_opt_hdr *)p; 2515 nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; 2516 nd_opt->nd_opt_len = len >> 3; 2517 lladdr = (char *)(nd_opt + 1); 2518 memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen); 2519 p += len; 2520 } 2521 LLE_RUNLOCK(ln); 2522 } 2523 nolladdropt:; 2524 2525 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2526 2527 /* just to be safe */ 2528 if (m0->m_flags & M_DECRYPTED) 2529 goto noredhdropt; 2530 if (p - (u_char *)ip6 > maxlen) 2531 goto noredhdropt; 2532 2533 { 2534 /* redirected header option */ 2535 int len; 2536 struct nd_opt_rd_hdr *nd_opt_rh; 2537 2538 /* 2539 * compute the maximum size for icmp6 redirect header option. 2540 * XXX room for auth header? 2541 */ 2542 len = maxlen - (p - (u_char *)ip6); 2543 len &= ~7; 2544 2545 /* 2546 * Redirected header option spec (RFC2461 4.6.3) talks nothing 2547 * about padding/truncate rule for the original IP packet. 2548 * From the discussion on IPv6imp in Feb 1999, 2549 * the consensus was: 2550 * - "attach as much as possible" is the goal 2551 * - pad if not aligned (original size can be guessed by 2552 * original ip6 header) 2553 * Following code adds the padding if it is simple enough, 2554 * and truncates if not. 2555 */ 2556 if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) { 2557 /* not enough room, truncate */ 2558 m_adj(m0, (len - sizeof(*nd_opt_rh)) - 2559 m0->m_pkthdr.len); 2560 } else { 2561 /* 2562 * enough room, truncate if not aligned. 2563 * we don't pad here for simplicity. 2564 */ 2565 size_t extra; 2566 2567 extra = m0->m_pkthdr.len % 8; 2568 if (extra) { 2569 /* truncate */ 2570 m_adj(m0, -extra); 2571 } 2572 len = m0->m_pkthdr.len + sizeof(*nd_opt_rh); 2573 } 2574 2575 nd_opt_rh = (struct nd_opt_rd_hdr *)p; 2576 memset(nd_opt_rh, 0, sizeof(*nd_opt_rh)); 2577 nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER; 2578 nd_opt_rh->nd_opt_rh_len = len >> 3; 2579 p += sizeof(*nd_opt_rh); 2580 m->m_pkthdr.len = m->m_len = p - (u_char *)ip6; 2581 2582 /* connect m0 to m */ 2583 m->m_pkthdr.len += m0->m_pkthdr.len; 2584 m_cat(m, m0); 2585 m0 = NULL; 2586 } 2587noredhdropt: 2588 if (m0) { 2589 m_freem(m0); 2590 m0 = NULL; 2591 } 2592 2593 /* XXX: clear embedded link IDs in the inner header */ 2594 in6_clearscope(&sip6->ip6_src); 2595 in6_clearscope(&sip6->ip6_dst); 2596 in6_clearscope(&nd_rd->nd_rd_target); 2597 in6_clearscope(&nd_rd->nd_rd_dst); 2598 2599 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 2600 2601 nd_rd->nd_rd_cksum = 0; 2602 nd_rd->nd_rd_cksum 2603 = in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen)); 2604 2605 /* send the packet to outside... */ 2606 if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0) 2607 icmp6_ifstat_inc(ifp, ifs6_out_error); 2608 2609 icmp6_ifstat_inc(ifp, ifs6_out_msg); 2610 icmp6_ifstat_inc(ifp, ifs6_out_redirect); 2611 ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT); 2612 2613 return; 2614 2615fail: 2616 if (m) 2617 m_freem(m); 2618 if (m0) 2619 m_freem(m0); 2620} 2621 2622/* 2623 * ICMPv6 socket option processing. 2624 */ 2625int 2626icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt) 2627{ 2628 int error = 0; 2629 struct in6pcb *in6p = sotoin6pcb(so); 2630 2631 if (sopt->sopt_level != IPPROTO_ICMPV6) 2632 return rip6_ctloutput(op, so, sopt); 2633 2634 switch (op) { 2635 case PRCO_SETOPT: 2636 switch (sopt->sopt_name) { 2637 case ICMP6_FILTER: 2638 { 2639 struct icmp6_filter fil; 2640 2641 error = sockopt_get(sopt, &fil, sizeof(fil)); 2642 if (error) 2643 break; 2644 memcpy(in6p->in6p_icmp6filt, &fil, 2645 sizeof(struct icmp6_filter)); 2646 error = 0; 2647 break; 2648 } 2649 2650 default: 2651 error = ENOPROTOOPT; 2652 break; 2653 } 2654 break; 2655 2656 case PRCO_GETOPT: 2657 switch (sopt->sopt_name) { 2658 case ICMP6_FILTER: 2659 { 2660 if (in6p->in6p_icmp6filt == NULL) { 2661 error = EINVAL; 2662 break; 2663 } 2664 error = sockopt_set(sopt, in6p->in6p_icmp6filt, 2665 sizeof(struct icmp6_filter)); 2666 break; 2667 } 2668 2669 default: 2670 error = ENOPROTOOPT; 2671 break; 2672 } 2673 break; 2674 } 2675 2676 return (error); 2677} 2678 2679/* 2680 * Perform rate limit check. 2681 * Returns 0 if it is okay to send the icmp6 packet. 2682 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate 2683 * limitation. 2684 * 2685 * XXX per-destination/type check necessary? 2686 */ 2687static int 2688icmp6_ratelimit( 2689 const struct in6_addr *dst, /* not used at this moment */ 2690 const int type, /* not used at this moment */ 2691 const int code) /* not used at this moment */ 2692{ 2693 int ret; 2694 2695 ret = 0; /* okay to send */ 2696 2697 /* PPS limit */ 2698 if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count, 2699 icmp6errppslim)) { 2700 /* The packet is subject to rate limit */ 2701 ret++; 2702 } 2703 2704 return ret; 2705} 2706 2707static struct rtentry * 2708icmp6_mtudisc_clone(struct sockaddr *dst) 2709{ 2710 struct rtentry *rt; 2711 int error; 2712 2713 rt = rtalloc1(dst, 1); 2714 if (rt == 0) 2715 return NULL; 2716 2717 /* If we didn't get a host route, allocate one */ 2718 if ((rt->rt_flags & RTF_HOST) == 0) { 2719 struct rtentry *nrt; 2720 2721 error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL, 2722 RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt); 2723 if (error) { 2724 rtfree(rt); 2725 return NULL; 2726 } 2727 nrt->rt_rmx = rt->rt_rmx; 2728 rtfree(rt); 2729 rt = nrt; 2730 } 2731 error = rt_timer_add(rt, icmp6_mtudisc_timeout, 2732 icmp6_mtudisc_timeout_q); 2733 if (error) { 2734 rtfree(rt); 2735 return NULL; 2736 } 2737 2738 return rt; /* caller need to call rtfree() */ 2739} 2740 2741static void 2742icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r) 2743{ 2744 2745 KASSERT(rt != NULL); 2746 rt_assert_referenced(rt); 2747 2748 if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) == 2749 (RTF_DYNAMIC | RTF_HOST)) { 2750 rtrequest(RTM_DELETE, rt_getkey(rt), 2751 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 2752 } else { 2753 if (!(rt->rt_rmx.rmx_locks & RTV_MTU)) 2754 rt->rt_rmx.rmx_mtu = 0; 2755 } 2756} 2757 2758static void 2759icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r) 2760{ 2761 2762 KASSERT(rt != NULL); 2763 rt_assert_referenced(rt); 2764 2765 if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) == 2766 (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) { 2767 rtrequest(RTM_DELETE, rt_getkey(rt), 2768 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 2769 } 2770} 2771 2772/* 2773 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes. silly? 2774 */ 2775static int 2776sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS) 2777{ 2778 (void)&name; 2779 (void)&l; 2780 (void)&oname; 2781 2782 if (namelen != 0) 2783 return (EINVAL); 2784 2785 return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp, 2786 /*XXXUNCONST*/ 2787 __UNCONST(newp), newlen)); 2788} 2789 2790static int 2791sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS) 2792{ 2793 2794 return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS)); 2795} 2796 2797static int 2798sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS) 2799{ 2800 int error, tmp; 2801 struct sysctlnode node; 2802 2803 node = *rnode; 2804 node.sysctl_data = &tmp; 2805 tmp = icmp6_redirtimeout; 2806 error = sysctl_lookup(SYSCTLFN_CALL(&node)); 2807 if (error || newp == NULL) 2808 return error; 2809 if (tmp < 0) 2810 return EINVAL; 2811 icmp6_redirtimeout = tmp; 2812 2813 if (icmp6_redirect_timeout_q != NULL) { 2814 if (icmp6_redirtimeout == 0) { 2815 rt_timer_queue_destroy(icmp6_redirect_timeout_q, 2816 true); 2817 } else { 2818 rt_timer_queue_change(icmp6_redirect_timeout_q, 2819 icmp6_redirtimeout); 2820 } 2821 } else if (icmp6_redirtimeout > 0) { 2822 icmp6_redirect_timeout_q = 2823 rt_timer_queue_create(icmp6_redirtimeout); 2824 } 2825 2826 return 0; 2827} 2828 2829static void 2830sysctl_net_inet6_icmp6_setup(struct sysctllog **clog) 2831{ 2832 extern int nd6_maxqueuelen; /* defined in nd6.c */ 2833 2834 sysctl_createv(clog, 0, NULL, NULL, 2835 CTLFLAG_PERMANENT, 2836 CTLTYPE_NODE, "inet6", NULL, 2837 NULL, 0, NULL, 0, 2838 CTL_NET, PF_INET6, CTL_EOL); 2839 sysctl_createv(clog, 0, NULL, NULL, 2840 CTLFLAG_PERMANENT, 2841 CTLTYPE_NODE, "icmp6", 2842 SYSCTL_DESCR("ICMPv6 related settings"), 2843 NULL, 0, NULL, 0, 2844 CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL); 2845 2846 sysctl_createv(clog, 0, NULL, NULL, 2847 CTLFLAG_PERMANENT, 2848 CTLTYPE_STRUCT, "stats", 2849 SYSCTL_DESCR("ICMPv6 transmission statistics"), 2850 sysctl_net_inet6_icmp6_stats, 0, NULL, 0, 2851 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2852 ICMPV6CTL_STATS, CTL_EOL); 2853 sysctl_createv(clog, 0, NULL, NULL, 2854 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2855 CTLTYPE_INT, "rediraccept", 2856 SYSCTL_DESCR("Accept and process redirect messages"), 2857 NULL, 0, &icmp6_rediraccept, 0, 2858 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2859 ICMPV6CTL_REDIRACCEPT, CTL_EOL); 2860 sysctl_createv(clog, 0, NULL, NULL, 2861 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2862 CTLTYPE_INT, "redirtimeout", 2863 SYSCTL_DESCR("Redirect generated route lifetime"), 2864 sysctl_net_inet6_icmp6_redirtimeout, 0, 2865 &icmp6_redirtimeout, 0, 2866 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2867 ICMPV6CTL_REDIRTIMEOUT, CTL_EOL); 2868#if 0 /* obsoleted */ 2869 sysctl_createv(clog, 0, NULL, NULL, 2870 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2871 CTLTYPE_INT, "errratelimit", NULL, 2872 NULL, 0, &icmp6_errratelimit, 0, 2873 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2874 ICMPV6CTL_ERRRATELIMIT, CTL_EOL); 2875#endif 2876 sysctl_createv(clog, 0, NULL, NULL, 2877 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2878 CTLTYPE_INT, "nd6_prune", 2879 SYSCTL_DESCR("Neighbor discovery prune interval"), 2880 NULL, 0, &nd6_prune, 0, 2881 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2882 ICMPV6CTL_ND6_PRUNE, CTL_EOL); 2883 sysctl_createv(clog, 0, NULL, NULL, 2884 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2885 CTLTYPE_INT, "nd6_delay", 2886 SYSCTL_DESCR("First probe delay time"), 2887 NULL, 0, &nd6_delay, 0, 2888 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2889 ICMPV6CTL_ND6_DELAY, CTL_EOL); 2890 sysctl_createv(clog, 0, NULL, NULL, 2891 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2892 CTLTYPE_INT, "nd6_umaxtries", 2893 SYSCTL_DESCR("Number of unicast discovery attempts"), 2894 NULL, 0, &nd6_umaxtries, 0, 2895 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2896 ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL); 2897 sysctl_createv(clog, 0, NULL, NULL, 2898 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2899 CTLTYPE_INT, "nd6_mmaxtries", 2900 SYSCTL_DESCR("Number of multicast discovery attempts"), 2901 NULL, 0, &nd6_mmaxtries, 0, 2902 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2903 ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL); 2904 sysctl_createv(clog, 0, NULL, NULL, 2905 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2906 CTLTYPE_INT, "nd6_useloopback", 2907 SYSCTL_DESCR("Use loopback interface for local traffic"), 2908 NULL, 0, &nd6_useloopback, 0, 2909 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2910 ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL); 2911#if 0 /* obsoleted */ 2912 sysctl_createv(clog, 0, NULL, NULL, 2913 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2914 CTLTYPE_INT, "nd6_proxyall", NULL, 2915 NULL, 0, &nd6_proxyall, 0, 2916 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2917 ICMPV6CTL_ND6_PROXYALL, CTL_EOL); 2918#endif 2919 sysctl_createv(clog, 0, NULL, NULL, 2920 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2921 CTLTYPE_INT, "nodeinfo", 2922 SYSCTL_DESCR("Respond to node information requests"), 2923 NULL, 0, &icmp6_nodeinfo, 0, 2924 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2925 ICMPV6CTL_NODEINFO, CTL_EOL); 2926 sysctl_createv(clog, 0, NULL, NULL, 2927 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2928 CTLTYPE_INT, "errppslimit", 2929 SYSCTL_DESCR("Maximum ICMP errors sent per second"), 2930 NULL, 0, &icmp6errppslim, 0, 2931 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2932 ICMPV6CTL_ERRPPSLIMIT, CTL_EOL); 2933 sysctl_createv(clog, 0, NULL, NULL, 2934 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2935 CTLTYPE_INT, "nd6_maxnudhint", 2936 SYSCTL_DESCR("Maximum neighbor unreachable hint count"), 2937 NULL, 0, &nd6_maxnudhint, 0, 2938 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2939 ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL); 2940 sysctl_createv(clog, 0, NULL, NULL, 2941 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2942 CTLTYPE_INT, "mtudisc_hiwat", 2943 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2944 NULL, 0, &icmp6_mtudisc_hiwat, 0, 2945 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2946 ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL); 2947 sysctl_createv(clog, 0, NULL, NULL, 2948 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2949 CTLTYPE_INT, "mtudisc_lowat", 2950 SYSCTL_DESCR("Low mark on MTU Discovery route timers"), 2951 NULL, 0, &icmp6_mtudisc_lowat, 0, 2952 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2953 ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL); 2954 sysctl_createv(clog, 0, NULL, NULL, 2955 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2956 CTLTYPE_INT, "nd6_debug", 2957 SYSCTL_DESCR("Enable neighbor discovery debug output"), 2958 NULL, 0, &nd6_debug, 0, 2959 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2960 ICMPV6CTL_ND6_DEBUG, CTL_EOL); 2961 sysctl_createv(clog, 0, NULL, NULL, 2962 CTLFLAG_PERMANENT, 2963 CTLTYPE_STRUCT, "nd6_drlist", 2964 SYSCTL_DESCR("Default router list"), 2965 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2966 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2967 ICMPV6CTL_ND6_DRLIST, CTL_EOL); 2968 sysctl_createv(clog, 0, NULL, NULL, 2969 CTLFLAG_PERMANENT, 2970 CTLTYPE_STRUCT, "nd6_prlist", 2971 SYSCTL_DESCR("Prefix list"), 2972 sysctl_net_inet6_icmp6_nd6, 0, NULL, 0, 2973 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2974 ICMPV6CTL_ND6_PRLIST, CTL_EOL); 2975 sysctl_createv(clog, 0, NULL, NULL, 2976 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2977 CTLTYPE_INT, "maxqueuelen", 2978 SYSCTL_DESCR("max packet queue len for a unresolved ND"), 2979 NULL, 1, &nd6_maxqueuelen, 0, 2980 CTL_NET, PF_INET6, IPPROTO_ICMPV6, 2981 ICMPV6CTL_ND6_MAXQLEN, CTL_EOL); 2982} 2983 2984void 2985icmp6_statinc(u_int stat) 2986{ 2987 2988 KASSERT(stat < ICMP6_NSTATS); 2989 ICMP6_STATINC(stat); 2990} 2991